Dynamic Coupling Strategy for Interdependent Network Systems Against Cascading Failures
I-Cheng Lin, Osman Yağan, Carlee Joe‐Wong
Abstract
In networked systems, initial failures at only a small part of the network may trigger a sequential failure process called cascading failures, which may eventually lead to the breakdown of the entire system. This vulnerability is further exacerbated in modern systems that consist of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">multiple</i> networks, each carrying a flow or a load, that are <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">interdependent</i> ; e.g., the smart-grid, transportation systems, etc. In this paper, we study the robustness of such systems against cascading failures under a flow-redistribution-based model where the flow from a failed node gets redistributed in part to nodes within the same network and in part to nodes in the other networks that are interdependent with the one where the initial failure took place. In such systems, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">coupling coefficients</i> between the networks that determine the portion of the flow that will distributed internally vs. externally are key to determining the robustness. Here, we investigate the progress of failures in a two-networked system and derive expressions that help characterize the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">final</i> fraction of surviving nodes in both networks as a function of the coupling coefficients. Then, we develop a new <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dynamic</i> coupling coefficient strategy that adjusts these parameters during the cascading failure process to improve the robustness. We show that our step-wise optimization (SWO) strategy significantly improves the robustness of the entire system compared to prior work that mainly focus on static coupling coefficients. This is shown through extensive simulations on different network topologies and different attack types.